Three-Dimensional Solution Structure and - American Chemical Society

Schools of Biological and Chemical Sciences, UniVersity of East Anglia, Norwich NR4 7TJ, U.K., Protein Structure Group,. Zeneca Pharmaceuticals, Alder...
0 downloads 0 Views 399KB Size
Biochemistry 1996, 35, 9505-9512

9505

Three-Dimensional Solution Structure and 13C Nuclear Magnetic Resonance Assignments of the Colicin E9 Immunity Protein Im9†,‡ Michael J. Osborne,§ Alexander L. Breeze,| Lu-Yun Lian,⊥ Ann Reilly,∇ Richard James,∇ Colin Kleanthous,∇ and Geoffrey R. Moore*,§ Schools of Biological and Chemical Sciences, UniVersity of East Anglia, Norwich NR4 7TJ, U.K., Protein Structure Group, Zeneca Pharmaceuticals, Alderley Park, Macclesfield, Cheshire SK10 4TG, U.K., and Biological NMR Centre, Leicester UniVersity, Leicester, U.K. ReceiVed February 20, 1996; ReVised Manuscript ReceiVed May 13, 1996X

ABSTRACT: The 86-amino acid colicin E9 immunity protein (Im9), which inhibits the DNase activity of colicin E9, has been overexpressed in Escherichia coli and isotopically enriched with 15N and 13C. Using the 3D CBCANH and CBCA(CO)NH experiments, we have almost completely assigned the backbone 13C resonances and extended previously reported 15N/1H backbone assignments [Osborne et al. (1994), Biochemistry 33, 12347-12355]. Side chain assignments for almost all residues were made using the 3D 13C HCCH-TOCSY experiment allied to previous 1H assignments. Sixty solution structures of Im9 were determined using the DIANA program on the basis of 1210 distance restraints and 56 dihedral angle restraints. The 30 lowest-energy structures were then subjected to a slow-cooling simulated annealing protocol using XPLOR and the 21 lowest-energy structures, satisfying the geometric restraints chosen for further analysis. The Im9 structure is well-defined except for the termini and two solvent-exposed loops between residues 28-32 and 57-64. The average RMSD about the average structure of residues 4-84 was 0.94 Å for all heavy atoms and 0.53 Å for backbone CR, CdO, and N atoms. The Im9 fold is novel and can be considered a distorted antiparallel four-helix bundle, in which the third helix is rather short, being terminated close to its N-terminal end by a proline at its C-terminus. The structure fits in well with available kinetic and biochemical data concerning the interaction between Im9 and its target DNase. Important residues of Im9 that govern specificity are located on the molecular surface in a region rich in negatively charged groups, consistent with the proposed electrostatically steered association [Wallis et al. (1995a), Biochemistry 34, 13743-13750].

Colicin E9 (Col E9) is a plasmid-encoded DNase that is secreted as part of the stress response system of some Escherichia coli (Luria & Suit, 1987; Eaton & James, 1989). It binds to the vitamin B12 extracellular receptor encoded by the btuB gene on target cells (Di Masi et al., 1973), and, following translocation across the outer and cytoplasmic membranes, its cytotoxic activity leads to cell death. To provide the host organism with immunity from the action of colicin E9, an inhibitor protein, Im9, is coordinately produced (Chak & James, 1986; James et al., 1987). This acidic protein binds to intact Col E9 and its DNase domain with a dissociation constant of ∼10-16 M, revealing the complex to be one of the tightest interprotein complexes known (Wallis et al., 1995a). It is the complex of Col E9 with Im9 that is released by the producing cell, but the complex is presumed to be dissociated during its interaction with the target cell since the DNase is active within it. In † This work was supported by the Wellcome Trust (a Research Leave Fellowship to G.R.M.) and by the Biotechnology & Biological Sciences Research Council. ‡ Coordinates have been deposited in the Brookhaven Protein Data Bank for the 21 final structures (1IMP) and the average structure (1IMQ). * Author to whom correspondence should be addressed. FAX: +441603 259396. E-mail: [email protected]. § School of Chemical Sciences, University of East Anglia. | Zeneca Pharmaceuticals. ⊥ Leicester University. ∇ School of Biological Sciences, University of East Anglia. X Abstract published in AdVance ACS Abstracts, June 15, 1996.

S0006-2960(96)00401-1 CCC: $12.00

addition to Col E9, colicin DNases E2, E7, and E8 have been isolated, all with their own specific inhibitor proteins: Im2, Im7, and Im8 respectively (Schaller & Nomura, 1976; Toba et al., 1988; Chak et al., 1991). There is a high degree of sequence similarity between the four colicin DNase domains (∼80%) and between their inhibitor proteins (∼50%), but despite this, only the cognate inhibitors provide complete immunity against the colicin DNase action (James et al., 1992; Wallis et al., 1995b). However, noncognate complexes are formed between the E9 DNase domain and Im2, Im7, and Im8 with Kd values ranging from 10-4-10-8M (Wallis et al., 1995b). The combination of high binding affinities between cognate partners and relatively low binding affinities between noncognate proteins, coupled with the relatively small size of the complex formed between a colicin DNase domain and an inhibitor protein (220 amino acids for Col E9/Im9), makes this an attractive system for investigating molecular recognition events between proteins. To pursue this, we are engaged in a detailed kinetic, protein engineering, and NMR study of the Col E9/Im9 proteins, both in their free states and in their interprotein complex. We have overexpressed the 134amino acid DNase domain of Col E9 (Wallis et al., 1994) and the 86-amino acid Im9 (Wallis et al., 1992a) and have used NMR spectroscopy to obtain structural information concerning both free and bound Im9 (Osborne et al., 1994). We have previously reported NMR studies on uniformaly 15N-labeled Im9 that provided assignments for 77 of its 86 © 1996 American Chemical Society

9506 Biochemistry, Vol. 35, No. 29, 1996 amino acid residues (Osborne et al., 1994). That study also showed that the protein contained at least three helical sections, with the specificity-determining region forming one of the helices. However, the incomplete assignments, coupled with severe overlap in the aliphatic region of the 1H-1H NOESY spectrum, prevented determination of a reliable three-dimensional (3D) structure at that stage, and to overcome these problems, we embarked on a study of uniformly 13C- and 15N-labeled Im9. In the present paper, we report further 1H and 15N assignments of Im9, together with 13C assignments, and describe its three-dimensional structure determined using NOE and backbone dihedral angle constraints obtained from multidimensional heteronuclear NMR data. MATERIALS AND METHODS Sample Preparation. Im9 was expressed in E. coli and purified as previously described (Wallis et al., 1992b). All samples prepared for NMR were assayed for biological activity as described by Wallis et al. (1992b) and found to be fully active. Uniformly 15N-labeled Im9 was obtained by growing the E. coli JM105 cells containing the expression system in minimal medium with 15NH4Cl (1 g/L), and uniformly 13C- and 15N-labeled Im9 was obtained from cells grown in minimal medium containing a mixture of [13C6]glucose (4 g/L) and 15NH4Cl (1 g/L). Yields of protein from these enrichments were typically 20 mg/L. Concentrations of NMR samples in either 90% 1H2O/10% 2H2O or 99.9% 2H2O were typically 1-4 mM. The NMR solutions were prepared from Im9 lyophilized from either 1H2O or 2H O solutions containing sodium phosphate at pH 7.0 or 2 6.2 to yield NMR samples with a final phosphate concentration of 25-50 mM. NMR Measurements. The NMR data were acquired on Bruker AMX600 or Varian Unity 600 spectrometers. Spectra measured with the Bruker instrument were recorded in the pure phase absorption mode by the time-proportional phase incrementation (TPPI) method (Redfield & Kunz, 1975; Marion & Wu¨thrich, 1983), and on the Varian instrument, quadrature detection was achieved in the indirectly acquired dimensions by the States-TPPI procedure (Marion et al., 1989b). Proton chemical shifts were measured from internal dioxane at 3.77 ppm; nitrogen chemical shifts were referenced using external 15NH4NO3 (5 M 15NH4NO3 in 2 M HNO3) with the ammonium 15N resonance taken as 0 ppm, and carbon chemical shifts were referenced to the methyl resonance of external sodium 3-(trimethylsilyl)propionate. Experimental details of the homonuclear and 1H-15N experiments are given in Osborne et al. (1994). NOESY (Jeener et al., 1979: Macura et al., 1981) data were collected using mixing times of 100 and 150 ms. Additional details of such experiments carried out for the present paper and details of the triple-resonance experiments are given below. All heteronuclear 3D experiments were carried out at 293 K. The 1H-15N 3D TOCSY-HMQC, NOESY-HMQC (Zuiderweg & Fesik, 1989; Marion et al., 1989a), and HMQC-NOESY-HMQC (Frenkiel et al., 1990) pulse schemes used were similar to those originally described. These were all measured on the Bruker instrument. The CBCANH and CBCA(CO)NH spectra were acquired with the Varian spectrometer using pulse schemes similar to those reported by Grzesiek and Bax (1992a,b, 1993). The data set sizes

Osborne et al. were 512 (1H) × 56 (13C) × 30 (15N) complex points with 16 scans recorded for the CBCA(CO)NH spectrum and 48 scans for the CBCANH spectrum. Water suppression was achieved with a 2.5 ms spin lock purge pulse. The CNNOESY-HSQC data were acquired with the pulse scheme of Pascal et al. (1994) on the Varian spectrometer, using pulsed-field gradients for water suppression and artifact reduction as described by Pascal et al. (1994). 13C(carbonyl) decoupling during t2 was accomplished using a cosinemodulated SEDUCE-1 decoupling waveform (McCoy & Mueller, 1992) applied off-resonance at the 13C(aliphatic/ aromatic) carrier, positioned at 67ppm. Acquisition times of 49 ms (t3,1H), 15.1 ms (t1, 1H), and 11.8 ms (t2, 15N/13C) were employed with mixing times of 100 and 150 ms. Data were processed on Silicon Graphics Indigo or Sun Sparc workstations using FELIX (Biosym/MSI). For the tripleresonance and CN-NOESY-HSQC data sets, conventional or “mirror image” (Zhu & Bax, 1990) linear prediction methods were employed to extend the data in the indirectly acquired dimensions where necessary. Structural Restraints and Calculations. A total of 1307 unique NOE cross-peaks were assigned from the 3D CNNOESY spectrum of a doubly labeled sample of Im9 recorded with a mixing time of 100 ms in 1H2O at 295 K. Peak volumes were translated into upper limits of interatomic distances using the CALIBA program and structures generated using DIANA, by the protocol described by Gu¨ntert et al. (1991). For the initial structure calculations, cross-peak volumes were calibrated from resonances arising from regions of the structure which were presumed to be relatively rigid (e.g. sequential amide proton connectivities at the center of R-helices). After the generation of initial structures using DIANA (Gu¨ntert et al., 1991), peak volumes were calibrated independently for the five calibration classes described by Gu¨ntert et al. (1991). To account for effects of spin diffusion and 13C relaxation times, a maximum upper distance limit of 6 Å was used. One hundred thirty-five of the distance restraints were deemed meaningless (i.e. the constraint defined a fixed distance or would not violate any possible conformation), leaving 1172 distance constraints from NOE data. Fifty-six 3JNHR coupling constants were measured from an HMQC-J experiment and converted to φ torsion restraints as follows. For 3JNHR of >8 Hz, φ was set to -120 ( 40°, and for 3JNHR values of O (2.3 ( 0.5 Å). No angle restraints were determined for β-methylene protons. Sixty structures were generated using the DIANA program. The 30 lowest-energy structures which had no common NOE-derived distance violations greater than 0.5 Å and no common torsion angle violations greater than 5° were then subjected to a slow-cooling simulated annealing protocol (initial temperature of 2000 K, 10 000 cooling steps) (Nilges et al., 1988) using the XPLOR program (Brunger, 1992). Twenty one of these structures satisfied all geometric restraints (no NOE >0.5 Å and no dihedral >5°), and from this ensemble, an average structure was computed and subjected to an additional 1000 cycles of Powell restrained energy minimization. Analysis of the structures was carried out with MOLMOL (Koradi et al., 1996) and QUANTA (Molecular Simulations).

Structure of Im9

FIGURE 1: Selected region of the 2D 1H-13C HSQC spectrum of resonance overlap was a problem.

Biochemistry, Vol. 35, No. 29, 1996 9507

13C-

and

15N-labeled

Im9. The boxed areas indicate regions where

RESULTS AND DISCUSSION Resolution of Resonances. Severe overlap of resonances in 2D 1H-1H spectra of Im9 necessitated the use of 3D 15Nedited experiments to obtain sequential assignments following the procedure of Wu¨thrich (1986). Using this approach, ∼93% of the backbone 1H resonances of Im9 were assigned and the secondary structure delineated on the basis of short range and medium range NOEs. These showed at least three helices and no β-sheet to be present in Im9 (Osborne et al., 1994). Highly helical proteins tend to display poor backbone proton chemical shift dispersion (Wishart et al., 1992), and this was also observed for Im9. Chemical shift degeneracy of, in particular, R and β protons and methyl resonances meant that only poor resolution structures of Im9 were obtained from experiments utilizing 15N-enriched samples of Im9. In general, overlap in the aliphatic region may be alleviated by editing spectra according to the 13C chemical shifts of the directly bound carbon nuclei, and once assignments of 13C chemical shifts are obtained, the often unique combination of 1H and 13C chemical shifts can be used to identify otherwise ambiguous NOEs in the 15N- edited 3D NOESY spectra (see later). The resolution afforded by the introduction of the 13C dimension for Im9 is evident from Figure 1 which shows a selected region of the 1H-13C HSQC spectrum of Im9. The majority of the 58 protons that resonate between 3.8 and 4.3 ppm in the 1H NMR spectrum are resolved, though there remain some areas, indicated by boxes in Figure 1, where overlap is still a problem. Assignment of these areas was achieved using tripleresonance experiments, with a 13C- and 15N-enriched sample, which make use of the sequential and intraresidue 1H and 15 N amide chemical shifts to separate the resonances. However, despite the assignments, ambiguity remains in assigning NOE cross-peaks from NOESY experiments, as described later. Assignment of Resonances. Assignments of the backbone 13C, 15N, and 1H resonances were obtained using a strategy similar to that of Grzesiek and Bax (1993). In this procedure, comparison of spectra from experiments which directly correlate peptide amide resonaces with the preceding interresidue 13CR and 13Cβ resonances [CBCA(CO)NH experiment] (Grzesiek & Bax, 1992b) and with both the intrares-

FIGURE 2: Region of a 600 MHz 13C-15N-1H CBCANH spectrum of uniformly 15N- and 13C-labeled Im9 in 90% 1H2O/10% 2H2O illustrating sequential assignments via heteronuclear scalar couplings. The figure shows 2D strips taken through the 3D CBCANH spectrum for residues Lys57-Ser63 in which the 13CR and 13Cβ chemical shifts of residue i are correlated with their intraresidue NH 15N and 1H chemical shifts and also the NH 15N and 1H chemical shifts of residue i + 1. Each strip represents a narrow D1(1H) region of a D2(13C)/D1(1H) cross section of the 3D spectrum at the NH 1H/15N frequency of an amino acid. Intraresidue 13CR and 13Cβ connectivities are marked. 13CR connectivities, except those of glycine, are positive (solid contours) and 13Cβ connectivities and glycine 13CR connectivities are negative (broken contours).

idue 13CR and 13Cβ resonances and the preceding interresidue and 13Cβ resonances (CBCANH experiment) (Grzesiek & Bax, 1992a) affords sequential assignment of residues. Amino acid type is then identified on the basis of the 13CR and 13Cβ chemical shifts. Comparison of the CBCANH spectrum (Figure 2) with the CBCA(CO)NH spectrum (not shown) confirmed the previously reported 1H and 15N assignments (Osborne et al., 1994) and allowed almost complete assignments for the 1H, 13C, and 15N backbone resonances of Im9 to be obtained. These assignments include those for the previously unassigned residues 62-64 (Figure 2) and 85-86 and the three proline residues, 47, 56, and 64. The 13Cβ shifts of the proline residues correspond to the random coil shifts for the trans configuration (Jardetzky & Roberts, 1981), which is in accordance with NOE analysis and the final solution 13CR

9508 Biochemistry, Vol. 35, No. 29, 1996

Osborne et al.

Table 1: Summary of NMR Parameters for Structure Calculations restraint type

number of restraints

intraresidue NOEs interresidue sequential NOEs (|i - j| ) 1) interresidue short range NOEs (1 < |i - j| < 5) interresidue long range NOEs (|i - j| g 5) total number of NOEs

358 311 269 234 1172

hydrogen bond restraints (one per H bond) φ dihedral angle restraints total number of restraints

38 56 1266

structure. The information from the triple-resonance experiments was important, particularly the CBCANH experiment which displayed both inter- and intraresidue connectivities for residues 3-86. The remaining side chain 13C and 1H assignments were identified from the 3D 13C-1H HCCH-TOCSY experiment in conjunction with previously assigned proton resonances (Osborne et al., 1994). This experiment afforded additional proton assignments, particularly for the glutamate and lysine side chains, and allowed a number of ambiguous assignments to be confirmed. The mixing time employed for this experiment, 20 ms, did not allow efficient carbon-carbon magnetization transfer between the CR and Cβ resonances of serine residues (Clore et al., 1990), and thus, unobservable, or weak, cross-peaks for these residues resulted. Additional serine proton resonances were assigned on the basis of their connectivities with 13CR and 13Cβ resonances detected in the CN-NOESY spectra. Assignment of NOEs. In our previous paper (Osborne et al., 1994), 3D 15N-edited NOESY-HMQC and HMQCNOESY-HMQC experiments were used to identify a large number of sequential and medium range NOEs involving amide and backbone protons and amide-amide protons. However, overlap in the aliphatic 1H region still caused assignment problems. For example, in the R-helices, there are a large number of degenerate HR and Hβ protons, making it difficult to distinguish between sequential and medium range RN, βN, or Rβ connectivities, which are important for defining helices (Wu¨thrich, 1986). More importantly, severe overlap often made it impossible to identify many long range NOE connectivities involving backbone atoms or methyl groups which also showed degeneracy in the 1H dimension. Indeed, from 15N-edited experiments, only 673 unambiguous NOE cross-peaks were identified, the majority of which were sequential or medium range and insufficient to define a high-resolution structure. We therefore utilized the 13C dimension as well by collecting a 3D CN-NOESYHSQC data set (Pascal et al., 1994). This experiment overcomes 1H chemical shift degeneracy according to the directly bound 15N or 13C nucleus by simultaneously detecting 13C and 15N frequencies in one dimension. Thus, protons attached to degenerate 13C or 15N shifts are clearly resolved. This experiment yields similar information to the 4D 15Nand 13C-edited NOESY experiment (Kay et al., 1990), but since only three dimensions are present, it affords better digital resolution and thus sensitivity. It was particularly useful for identifying NOEs involving amide and backbone protons which were previously ambiguous and in identifying NOE connectivities involving methyl resonances. This experiment helped unambiguously assign a further 634 unique NOE cross-peaks, making a total of 1307 assigned NOE cross-peaks, of which 1172 were used for structure determination (Table 1). In regions of the spectrum where

FIGURE 3: Selected 2D 1H-1H slices taken from a 150 ms 3D CN-NOESY-HSQC data set of 13C- and 15N-labeled Im9 in 90% 1H O/10% 2H O. The slices are at F frequencies corresponding to 2 2 2 the 15N chemical shift of the side chain NH protons of Asn 24 (a 13 and b) and the C chemical shift of the γ-methyl protons of Thr20 (c). The diagonal cross-peaks are indicated by an asterisk, and the symmetry-related cross-peaks, showing NOEs between these protons, connected by a horizontal line.

overlap persists and was not resolved by the CN-NOESY experiment (for examples, see Figure 1), initial structure calculations were used to remove ambiguity. This involved structure determination using DIANA leaving out some of the NOESY cross-peak data and the derived structure used to assign unambiguous NOESY cross-peaks. The distance constraints table was then revised to include the new assignments, and structure calculations to assist assignment were again carried out. This procedure was repeated until no further ambiguous NOESY cross-peaks could be assigned. There remained NOESY cross-peaks where firm assignments could not be made even with the aid of the structure, and therefore, these NOEs were not used in the calculations. These included, for example, many NOE connectivities involving the R-carbons of residues His5 and Ser6 and the R-carbons of residues 30-33. An example of the excellent resolution afforded by the 3D CN-NOESY-HSQC experiment is shown in Figure 3 to illustrate a set of nonsequential NOE connectivities. Chemical Shift Analysis. The 13CR and 13Cβ chemical shifts can be used to identify amino acid type and are exquisitely sensitive to secondary structure in folded proteins. The chemical shift index (CSI) proposed by Wishart et al. (1992) and Wishart and Sykes (1994) can predict protein secondary structure on the basis of analyses of 1HR, 13CR, and 13Cβ chemical shifts relative to their random coil values with a predictive accuracy in excess of 92%. 13CR resonances for residues in helices tend to be shifted >0.7 ppm to higher frequency than their random coil values. For β-sheets, perturbations of >0.7 ppm to lower frequency are usually seen for 13CR peaks. 1HR resonances in helices tend to exhibit

Structure of Im9 perturbations to lower frequency (>1.0 ppm) and vice versa for β-sheets. Continuous stretches of these perturbations are used to delineate secondary structure. The consensus CSI for Im9 is given in Figure 4a. This analysis indicates Im9 contains four R-helices involving residues 11-23, 28-42, 48-52, and 65-77. Such an analysis agrees well with the secondary structure of Im9 determined from the DIANA/ XPLOR structure calculations (Figure 5). Structural Restraints and Statistics. The structural statistics for the 21 converged structures (Figure 6) calculated from the structural restraints given in Table 1 are reported in Table 2. The final 21 structures exhibit an atomic RMSD about the average of 0.64 Å for the backbone CR, CdO, and N atoms and 1.12 Å for all non-hydrogen atoms, indicating that the structures are reasonably well-defined by the experimental restraints. Additionally, deviations from idealized covalent geometry are very small, indicating that no severe distortions exist in the 21 structures. A more detailed assessment of local variability of atoms is provided by a plot of RMSD versus residue number (Figure 4c). This figure shows that the majority of the backbone atoms are welldefined, in particular the four helices whose backbone atoms exhibit an RMSD of 0.28 Å. Backbone atoms which are less well-defined are the termini and segments comprising residues 27-31 (which still exhibit RMSDs below 1 Å) and residues 56-62. These residues are relatively poorly defined by the experimental restraints (Figure 4d) and are also found at the surface of the protein (Figure 4b). Leaving out the terminal residues from the comparison of 21 structures indicates that the core structures are relatively well-defined: atomic RMSDs for residues 4-84 about the average are 0.53 Å for the backbone CR, CdO, and N atoms and 0.94 Å for all non-hydrogen atoms. The local RMSD for the side chains is also well-defined, with ∼70% showing RMSD values below 1 Å. Residues whose side chains exhibit RMSDs above 1 Å are mainly located at the surface and are therefore likely to exhibit multiple conformations. The lack of NOE data at the termini and in the two loop regions may reflect fluctuations in these areas, and this point is addressed later. The Ramachandran plot (Figure 7) for the restrained energy-minimized average structure further indicates the quality of the Im9 structure, with the majority of residues falling in the allowed regions of the plot. Residues 3, 4, 55, 60, 80, and 84 are outside the allowed regions. No residue falls into a disallowed region in all 21 experimental structures, and with the exception of residue 55, the residues with unfavorable dihedral angles in the average structure are located in loop regions or close to the termini and thus have correspondingly high RMSD values. Similar behavior was observed for some of the loop and terminal residues in the solution structure of barstar (Lubienski et al., 1994). Residues exhibiting positive φ values (25, 29, 45, 49, 59, 79, and 82) also appear in loop regions. This has been noted in other proteins and may reflect how energetically unfavorable φ angles allow short loops to connect secondary structural elements (Ludvigsen & Poulsen, 1992; Serrano et al., 1992; Lubienski et al., 1994). Description of the Polypeptide Fold of Im9. The structure of Im9 (Figures 5 and 6) consists of three well-defined R-helicesshelix I (Glu12-Cys23), helix II (Glu30-Thr44), and helix IV (Ser65-A77)swith an additional one-turn 310 helix (Asp51-Tyr55). Helices I and II are antiparallel (forming an angle of -150°) and are connected by a loop comprising six residues which is relatively ill-defined. This

Biochemistry, Vol. 35, No. 29, 1996 9509

FIGURE 4: (a) Histograms showing the consensus chemical shift index of Im9 on the basis of the 13C and 1H chemical shifts (Wishart et al., 1992; Wishart & Sykes, 1994). (b) Relative surface accessibility per residue for a sphere with a 1.4 Å radius. (c) Distribution of the atomic RMSDs about the mean structure for the 21 solution structures of Im9 as a function of residue number for the CR, CdO, and N backbone atoms (solid) and all nonhydrogen atoms (open). (d) Number of interresidue (shaded) and intraresidue (solid) NOEs per amino acid residue.

turn is highly solvent accessible as indicated by Figure 4b and NH exchange data; exchange throughout this stretch occurs within an 8 min period prior to recording of the first data set (Osborne et al., 1994). Amide exchange rates in

9510 Biochemistry, Vol. 35, No. 29, 1996

Osborne et al.

FIGURE 5: Stereoview showing the superposition of the CR, CdO, and N backbone atoms for the 21 solution structures of Im9.

FIGURE 6: Ribbon structure of the energy minimized average structure of Im9 generated using the program MOLSCRIPT (Kraulis, 1991). Table 2: Summary of Structure Statistics RMS Deviation from Experimental Restraints distance 0.03 ( 0.001 Å dihedral angle 0.43 ( 0.20° Deviations from Ideal Covalent Geometry Based on the XPLOR Force Field bonds 4.97 × 10-3 ( 0.237 × 10-3 Å angles 0.65 ( 0.02° impropers 0.44 ( 0.02° Atomic RMS Differences from Average Structure backbone atoms 0.64 Å all heavy atoms 1.12 Å

helix I tend to be the longest of the whole molecule (Osborne et al., 1994) and may reflect stabilization caused by a large number of contacts with helices II and IV and, to a lesser extent, helix III. Helix II constitutes the longest helix in Im9. It is welldefined by the NMR data and forms contacts with the three other helices, but primarily with helix I. A well-defined β-turn between residues 45 and 49 connects helices II and III at an angle of 100° to each other. Again, positive φ angles

FIGURE 7: Ramachandran plot for the energy-minimized average structure of Im9 generated by MOLMOL (Koradi et al., 1996). The broken lines indicate the 95% allowed regions; + indicates glycine residue; and closed circles indicate other residues. Leu3, Tyr55, Asp60, Lys80, and Lys84 fall outside the 95% allowed regions (see text).

are observed for members of this connecting unit (Glu45 and Gly49). Despite its small size, helix III is well-defined by the NMR data and forms many contacts with the other helices. Additionally, the amide protons exchange relatively slowly (Osborne et al., 1994), although one might have expected them to be relatively rapidly exchanging since it is only a one-turn helix. The presence of bulky hydrophobic groups at the surface of this helix might protect the NH protons from the solvent. Helices III and IV are connected by a long ill-defined loop incorporating residues 57-64, four of which have carboxylate side chains. Amide exchange data (Osborne et al., 1994) and surface accessibility calculations of the minimized average structure (Figure 4b) show that this loop, and many of its side chain atoms, are accessible to solvent. The poor definition in this area reflects a lack of restraining interresidue interactions in keeping with this loop being relatively

Structure of Im9 unconstrained in solution. This is consistent with the CSI analysis which indicates residues in this loop are in random configurations (Figure 4a). The final helix brings both termini close together to allow an H bond between the NH of Lys84 and the backbone O of Tyr10 which is reflected in amide exchange data. In addition to contacts with the other helices, helix IV has contacts with the loop linking helices II and III, with Trp74 particularly close to His46. Side Chains. About 70% of the side chains are welldefined. Continuous stretches with RMSDs of 1.2 Å or greater are observed for the N- and C-termini and the loop regions involving residues 28-32 and 57-64. The side chains of these residues are situated at, or close to, the surface of Im9 and are therefore highly solvent accessible (Figure 4b) and in environments where they have relatively unrestricted motion compared to other regions of Im9. The large RMSDs for these loops may therefore reflect mobility, and this may also account for the small number of NOEs observed for these regions (Figure 4d). Line widths of peptide NH resonances in 1H-15N HSQC spectra do not vary significantly along the peptide chain, showing that any differential mobility is not on a time scale to affect these. The origin of the large RMSDs in the 28-32 and 57-64 loops is currently under investigation with variant proteins produced by site-directed mutagenesis and relaxation time measurements. The core of Im9 is stabilized by a large number of interactions involving hydrophobic side chains which, with the exception of some of the buried aromatic residues, are relatively well-defined. Figure 4c shows that the RMSDs of the side chains of Tyr10, Phe15, Phe40, and Phe83 are 1.0-1.8 Å. In addition to this, their respective Hδ and H chemical shifts are degenerate. This indicates that these residues flip rapidly about their Cβ-Cγ bonds (Wu¨thrich, 1986), though the necessary structural fluctuations required to permit this are not fully reflected in the ensemble of structures shown in Figure 5. This situation has been observed for the majority of protein solution structures and indicates that an ensemble of NMR structures do not reflect all large scale amplitudes in proteins. On the basis of fractional surface accessibility calculations, with a value of 10, has revealed that the rate of association is very salt dependent and in buffers of low ionic strength reaches a limiting value of 4 × 109 M-1 s-1 (Wallis et al., 1995a). These data indicate that there is an electrostatically steered collision between the two proteins to form an encounter complex which then undergoes a structural rearrangement to yield the final stable structure. This model implies that the DNase binding site of Im9, which is likely to include helices II and III, should be rich in acidic groups, and this is indeed the case. Glu30, Glu31, Glu32, Glu41, Asp45, and Asp51 are all part of this surface. By contrast, there is only one basic group in this region, Lys35. Mutagenesis studies, currently underway, are required to delineate further the interaction region on Im9. ACKNOWLEDGMENT We gratefully acknowledge Prof. K. Wu¨thrich for making available to us the program DIANA. We also thank Dr. R. Boetzel (UEA) for her assistance with the Ramachandran analysis. SUPPORTING INFORMATION AVAILABLE A virtually complete list of 1H, 13C, and 15N assignments for Im9 (7 pages). Ordering information is available on any current masthead page. REFERENCES Chak, K.-F., & James, R. (1986) J. Gen. Microbiol. 132, 61-71. Chak, K.-F., Kuo, W.-S., Lu, F.-M., & James, R. (1991) J. Gen. Microbiol. 137, 91-100. Clore, G. M., Bax, A., Driscoll, P. C., Wingfield, P. T., & Gronenborn, A. M. (1990) Biochemistry 29, 8172-8184. Di Masi, D. R., White, D. C., Schnaitman, C. A., & Bradbeer, C. (1973) J. Bacteriol. 115, 506-513. Eaton, T., & James, R. (1989) Nucleic Acids Res. 17, 1761-1761. Frenkiel, T., Bauer, C., Carr, M. D., Birdsall, B., & Feeney, J. (1990) J. Magn. Reson. 90, 420-425. Grzesiek, S., & Bax, A. (1992a) J. Magn. Reson. 99, 201-207. Grzesiek, S., & Bax, A. (1992b) J. Am. Chem. Soc. 114, 62916293. Grzesiek, S., & Bax, A. (1993) J. Biomol. NMR 3, 185-204. Guillet, V., Lapthorn, A., Hartley, R. W., & Maugen, Y. (1993) Structure 1, 165-176. Gu¨ntert, P., Braun, W., & Wu¨thrich, K. (1991) J. Mol. Biol. 217, 517-530. Holm, L., & Sander, C. (1993) J. Mol. Biol. 223, 123-138. Holm, L., & Sander, C. (1994) Nucleic Acids Res. 22, 3600-3609. James, R., Curtis, M. D., Wallis, R., Osborne, M. J., Kleanthous, C., & Moore, G. R. (1992) in Bacteriocins, microcins & lantibiotics (James, R., Lazdunski, C., & Pattus, F., Eds.) pp 181-201, NATO ASI Series H, Springer, Heidelberg. Jardetzky, O., & Roberts, G. C. K. (1981) NMR in Molecular Biology, p 153, Academic Press, New York. Jeener, J., Meier, B. H., Bachmann, P., & Ernst, R. R. (1979) J. Chem. Phys 71, 4546-4553.

Osborne et al. Kay, L. E., Clore, G. M., Bax, A., & Gronenborn, A. M. (1990) Science 249, 411. Koradi, R., Billeter, M., & Wu¨thrich, K. (1996) J. Mol. Graphics (in press). Kraulis, P. J. (1991) J. Appl. Crystallogr. 24, 946-950. Lubienski, M. J., Bycroft, M., Freund, S. M. V., & Fresht, A. R. (1994) Biochemistry 33, 8866-8877. Ludvigsen, S., & Poulsen, F. M. (1992) J. Biolmol. NMR 2, 227233. Luria, S. E., & Suit, J. L. (1987) in Escherichia coli & Salmonella typhimurium, cellular & molecular biology (Neidhardt, F. C., Ed.) Vol. 2, pp 1615-1624, American Society for Microbiology, Washington, DC. Macura, S., Huang, Y., Suter, D., & Ernst, R. R. (1981) J. Magn. Reson. 43, 259-281. Marion, D., & Wu¨thrich, K. (1983) Biochem. Biophys. Res. Commun. 113, 967-974. Marion, D., Driscoll, P,. C., Kay, L. E., Wingfield, P. T., Bax, A., Gronenborn, A. M., & Clore, G. M. (1989a) Biochemistry 28, 6150-6156. Marion, D., Ikura, M., Tschudin, R., & Bax, A. (1989b) J. Magn. Reson. 85, 393-399. McCoy, M. A., & Mueller, L. (1992) J. Am. Chem. Soc. 114, 21082112. Nilges, M., Clore, G. M., & Gronenborn, A. M. (1988) FEBS Lett. 229, 317-324. Osborne, M. J., Lian, L.-Y., Wallis, R., Reilly, A., James, R., Kleanthous, C., & Moore, G. R. (1994) Biochemistry 33, 1234712355. Pascal, S. M., Muhandiram, D. R., Yamazaki, T., Forman-Kay, J. D., & Kay, L. E. (1994) J. Magn. Reson. B103, 197-201. Redfield, A. G., & Kunz, S. D. (1975) J. Magn. Reson. 19, 250254. Schaller, K., & Nomura, M. (1976) Proc. Natl. Acad. Sci. U.S.A. 73, 3989-3993. Serrano, L., Sancho, J., Hirshberg, M., & Fersht, A. R. (1992) J. Mol. Biol. 227, 544-559. Toba, M., Masaki, H., & Ohta, T. (1988) J. Bacteriol. 170, 32373242. Wallis, R., Moore, G. R., Kleanthous, C., & James, R. (1992a) Eur. J. Biochem. 210, 923-930. Wallis, R., Reilly, A., Rowe, A., Moore, G. R., James, R., & Kleanthous, C. (1992b) Eur. J. Biochem. 207, 687-695. Wallis, R., Reilly, A., Barnes, K., Abell, C., Campbell, D. G., Moore, G. R., James, R., & Kleanthous, C. (1994) Eur. J. Biochem. 220, 447-454. Wallis, R., Moore, G. R., James, R., & Kleanthous, C. (1995a) Biochemistry 34, 13743-13750. Wallis, R., Leung, K.-Y., Pommer, A. J., Videler, H., Moore, G. R., James, R., & Kleanthous, C. (1995b) Biochemistry 34, 13751-13759. Wishart, D. S. & Sykes, B. D. (1994) J. Biomol. NMR 4, 171180. Wishart, D. S., Sykes, B. D., & Richards, F. M. (1992) Biochemistry 31, 1647-1651. Wu¨thrich, K. (1986) in NMR of Proteins and Nucleic Acids, Wiley, New York. Yajima, S., Muto, Y., Morikawa, S., Nakamura, H., Yokoyama, S., Masaki, H., & Uozumi, T. (1993) FEBS Lett. 333, 257260. Zhu, G., & Bax, A. (1990) J. Magn. Reson. 90, 405-410. Zuiderweg, E. R. P., & Fesik, S. (1989) Biochemistry 28, 2387-2391. BI960401K